This prospective, longitudinal study will examine the clinical problem of symptomatic episodes that may present with sustained fever in patients with documented human immunodeficiency virus (HIV) infection. To contribute to the development of a comprehensive strategy for the diagnostic evaluation of these patients, the study will: 1) describe the spectrum, frequency and outcome of symptomatic episodes, 2) estimate the effects of medical procedures, fever and baseline health-related quality of life on ability to make diagnosis, mortality and subsequent health related quality of life, 3) characterize distinct clinical syndromes that allow a more focused investigation, 4) describe the yield of commonly used diagnostic tests, 5) identify patient characteristics that support either empiric therapy, intensive investigation or no therapy, and 6) develop predictive models for the yield of different diagnostic strategies and, where appropriate, diagnostic algorithms. The study will be conducted in a socioeconomically and racially diverse population, including at least 600 inpatients and outpatients at a county hospital (50%), and patients in four large practices devoted to the care of HIV-infected patients. (50%). Criteria for eligibility are: 1) documented positive HIV infection, or diagnosis of acquired immunodeficiency syndrome (AIDS), 2) ate > 18, and 3) either a) pulmonary symptoms, b) diarrhea or c) constitutional symptoms i) lasting at least 7 days or ii) resulting in initiation of specific therapy for a documented disorder. Patients will be enrolled at the time of documentation of a symptomatic episode and will be followed for no less than 12 months (unless death intervenes) and for up to 32 months. Using medical record review and patient interviews, and applying validated instruments to measure overall health status, the study will examine the usefulness of diagnostic investigation at different stages of the illness - overall and in subpopulations with the three clinical syndromes each with and without fever. The predictive models and algorithms that result from this study should allow efficient and appropriate investigation of patients with symptomatic illness with or without sustained fever, thereby avoiding unnecessary hospitalizations and excessive testing, and maximizing benefit to the patient.